Update: mt32emu-qt offers a possibility to play those demo songs nowadays. Apparently, working out the format was very easy, but grabbing songs from ROM is not, in fact. The first thing to note: these songs need special voice and rhythm patches that are unavailable otherwise (presumably those were ripped from another thing like D-10, not sure). Yet, I dunno how legal it would be to rip that data from the ROM, but playing the songs with mt32emu-qt should be perfectly legal, I hope 😀

Oh, and good news: I suddenly tried to record MIDI with mt32emu-qt playing a demo song, and it did the job. Weirdly, I thought it wouldn't for some reason, it was quite some time ago 😒

So one can actually do the conversion with mt32emu-qt, just need a new-gen ROM and some time...

I also thought it wouldn't! Will look into this, thanks very much!

@sergm

Gave this a try - it does seem you can record demos, sort of.

Problems I'm having:

- There's obviously some sort of patch being applied that isn't captured, which is very necessary for proper playback
- If you allow a demo song to play through to the end, you aren't prompted to save the capture, nor is it kept

Any suggestions?

Further update - got a bit further, but my captures are still bad.

Now I'm getting the patches in the SysEx stream, and they are played back on my MT-32...but I get "Exc. Checksum error" on playback just before the notes begin to play and they sound wrong (because the patch is failing to upload properly).

ANY help appreciated =) So close!

FIgured out the checksum error...it really is an invalid patch SysEx that is padded with FF instead of 00.

Got that fixed...now some songs play more or less OK and some sound awful due to bad patches, but I can't work out why the patches are bad...they are EXACTLY as recorded. Ugh.

The attached is 1:1 recording of the first demo song, direct from MUNT into straight MIDI, with the SysEx patch for Hi Tom 1 corrected (FF->00 and checksum fixed).

It sounds awful. I don't understand why.

That's odd. I did the same and succeeded. Note, munt is agnostic to the delay between SysEx messages, unlike the old-gen units, so I would expect the need to edit the MIDI stream to make it a product. Indeed, the recording has to be started prior to playing a song: pretty much each of them relies on custom patches, as I said. And hence, we want to cut the leading delay, at the very least. But we likely also want to _increase_ delays between SysEx messages, so old-gen units might process them without errors, I gather.

BTW, my captured MIDI file is much shorter, only 13 Kb 😀 Do you use the same Record button? Or an external sequencer?

sergm wrote on 2023-10-14, 08:53:

BTW, my captured MIDI file is much shorter, only 13 Kb 😀 Do you use the same Record button? Or an external sequencer?

Ah, sorted it. Please, switch off the test MIDI driver - it has no use here. Instead, tick the Pin checkbox, so the synth remains active in between MIDI sessions.

OK, I guess it'd be easiest to use a DOS MIDI player from modern DOSBox yet enable the delaysysex thing like in

1midiconfig=2 delaysysex

This will automagically delay all the SysEx messages to be compatible with old-gen devices.

Hmm, well. These songs are seemingly only for new-gen units. They really play odd on old devices even when we ensure delays between SysExes.

Oh, right. The SysEx that conveys the patch map is way too long for old-gen units. This needs to be split...

OK, that sounds like it explains the oddness I'm seeing on old-gen...but I'm afraid splitting the SysEx messages is likely beyond my capabilities.

For convenience of editing, I've moved all SysEx messages to a separate file. I have also replaced the FF bytes with zeros and split the last longish SysEx. Seems to sound more-or-less OK now.

From what I've been able to tell, the instrument patch sysex is identical across all the demos, is that correct?

IIRC, either the patch maps or the rhythm maps are different, this stuff prepends every song.

To be clear, I attached the "fixed" files for consideration only. Indeed, mt32emu-qt can deal with the illegal bytes yet ensure that all SysExes fit into the expected size 😀

Right, that's what I mean...for doing the remaining four, if I do the capture and pull out the sysex (and fix the display stuff for each song), I should be able to reuse your sysex for all the MIDs, correct?